GPT-6.2 Astra tops Blueprint-Bench 2, beating the human baseline on 3D floorplans
i_dg23 · x · 2026-11-20
Andon Labs' Blueprint-Bench 2 tests physical 3D understanding by having AI agents draw floorplans from photographs of apartment interiors. GPT-6.2 Astra ranks #1, with GPT-6.2 Sol at #3.
Per the score chart, only GPT-6.2 Astra and Gemini 4 Carbon clear the human baseline of 0.59; most frontier models cluster between 0.2 and 0.55, with Kimi K2.6 near the bottom.
More from Models
- "Why is nobody talking about Neon?" — mystery model draws hype from big names — RubenEVillegas · 2026-11-20
- Opus 3 as agent safety reviewer is too strict and paranoid, dev finds — MikePFrank · 2026-11-20
- 220B MoE with full 262K context runs on RTX 3090s: 251 tok/s on 4 cards — QuixiAI · 2026-11-20
- Suspected GPT-6 Astra leak: one-prompt iterative 3D scene edits with editable Blender files — ZeroStateReflex · 2026-11-20
- Gemini Ultra's 20x tier gives 4x the quota of 5x — Codex and Claude Code only double it — mazzaTalk · 2026-11-20
- Viral Claude-vs-ChatGPT Download Chart Was Wrong: Ratings Mistaken for Downloads — tinyfool · 2026-11-20